Magnetic penetration depth measurements of superconducting thin films by a microstrip resonator technique

Transcription

1 Magnetic penetration depth measurements of superconducting thin films by a microstrip resonator technique B. W. Langley,a) S. M. Anlage,b) R. F. W. Pease,a) and M. R. Beasley Department of Applied Physics, Stanford University, Stanford, California (Received 1 March 1991; accepted for publication 2 April 1991) The microstrip resonator technique is a convenient way to sensitively measure the temperature dependence of the magnetic penetration depth, n(t), in superconducting thin films. Because the method relies on measuring the resonant frequency of a microwave transmission line resonator, one can very precisely measure small changes in A( T). This technique can resolve changes in,4 on the order of several angstroms, allowing a direct measurement of the low-temperature behavior of A( T), which is a measure of the low-lying pair breaking excitations of the superconductor. Absolute penetration depth values can also be obtained from a self-consistent fit to the data to an assumed temperature dependence. Measurements of the penetration depth of Nb and NbCN film give results that are consistent with the predictions of BCS theory in which 2A/kT, is treated as an adjustable parameter, while YBa2Cu films give results that are not completely understood at this time. We also compare this technique with other methods of measuring the penetration depth of superconducting thin films, and discuss the systematic errors present in the measurement. 1. INTRODUCTION In response to an external magnetic field, a superconductor generates currents near its surface that shield out the applied field. The characteristic depth to which the shielding currents flow is the magnetic penetration depth il. This penetration depth is temperature dependent, starting from a finite value, &, at T = 0 K, and increasing monotonically as T approaches T, until the normal state skin depth dominates the electromagnetic response. The penetration depth is sensitive to the properties of the superconductor near its surface, or in a film of thickness t <A, to the entire film. The temperature dependence of the penetration depth is also sensitive to the nature of the quasiparticle excitation spectrum, since the penetration depth increases as the superconducting carrier density is depleted by thermal excitations. For instance, a superconductor with a nonzero energy gap over its entire Fermi surface will have an exponentially activated penetration depth temperature dependence at low temperatures, which is determined by the Boltzmann distribution of quasiparticles over its lowest energy gap. On the other hand, a superconductor with nodes in the energy gap will show a polynomial dependence for,4( T) at low temperatures since there is a continuous spectrum of quasiparticle excitations. To distinguish between these two possibilities it is very important to determine the low-temperature dependence of /2(T) as accurately as possible. Also, at all temperatures l/l2 is proportional to the square of the modulus of the superconducting order parameter, whose temperature dependence is sensitive to the magnitude of the gap 2A( O)/kT, Novel superconducting materials such as the heavy fermion, organic, and possibly copper-oxide classes are thought to have nonconventional pairing symmetry, which could lead to a distinctive non-bcs temperature dependence of A(T). Thus, careful measurements of ;1 allows one to investigate the pairing state of such materials.2 Many methods have been used to measure the magnetic penetration depth in superconducting thin films. Naturally, most of these method rely upon measuring the diamagnetic response of the superconductor. One of the most direct techniques involves measuring the magnetization of the sample as a function of temperature and applied field. In the normal state, a steady magnetic field will completely permeate a metal since the screening currents will eventually die out. In the superconducting state, however, the shielding currents will persist and the static magnetic field will be shielded out over a characteristic length il, giving rise to a diamagnetic moment. Measurements of the diamagnetic susceptibility xdc = M/H of a superconductor as a function of temperature, and knowledge of the sample geometry, allow one to calculate changes in /2. Absolute values of ;1 can then be obtained with the use of a theoretical, or independently measured, penetration depth temperature dependence /2( T)/A( O).3 The advantage of this method lies in the fact that it is a direct measurement of the magnetic properties of the material. On the other hand, the detailed morphology and dimensions of the sample must be known in order to calculate the magnitude of the penetration depth. Single crystal samples with a simple geometry prove to be the most reliable way of determining the magnetic penetration depth by magnetization.4 One can also determine il by measuring the ac susceptibility xac = dm/dh obtained when an alternating magnetic field is applied to the sample. The ac susceptibility is Also at Stanford Electronics Laboratory, Stanford University. b)present address: Physics Department, Center for Superconductivity Research, University of Maryland, College Park, Maryland Rev. Sci. Instrum. 62 (7), July American Institute of Physics 1601

2 simply related to the magnetization of the sample, and similar considerations as above apply to the extraction of/z from the data. The measurement is often done by placing one or more pickup coils around the sample and measuring their self or mutual inductance. This technique was originally employed by Schawlow and Devlin to study the penetration depth in superconducting tin.5 Sridhar, Wu, and Kennedy6 have used this technique to measure the magnetic field dependence of ;1 in single crystal YBa2Cu30, _ & Another method that makes use of the inductance of a superconductor is the mutual inductance technique. A superconducting film is positioned such that it modifies the mutual inductance of a pair of coils. The coils are wound astatically to confine the magnetic field and eliminate the effects of stray external field anrd pickup. One coil is driven at kilohertz frequencies and th.e pickup voltage in the receive coil is monitored by phase-sensitive detection. The in-phase and quadrature signals can be analyzed to yield the transverse penetration depth R2/d (d is the film thickness) and the dissipation. The technique has been used to study the penetration depth in high-t, thin fdms and the Kosterlitz-Thouless transition in type-ii superconductors9 Quantitative interpretation of the data relies upon a calculation of the current distribution induced in the film by the drive coi1,8 as well as knowledge of the coil dimensions and spacings. The film must be thin compared to R and very homogeneous to measure the low temperature dependence of /2. Flux pinning can also affect the result. Another measurement tech.nique, which has traditionally been used to study the surface impedance of superconductors at microwave frequencies, is the resonant cavity. Cavity walls can be coated with a superconductor to a few penetration depths in thickness, or the entire cavity can be made from the material of interest. Alternatively, a superconducting thin film can be used as an end plate for a cylindrical or rectangular cavii:y,12 or the superconductor can be placed in a high rf magnetic field region inside the cavity.13, 4 The resonant freqtrency of the cavity is influenced by the penetration depth of the superconductor through the cavity inductance. Also, the Q of the cavity resonance is influenced by the superconducting losses. Thus the shift in resonant frequency of the cavity as a function of temperature can be used to determine changes in the penetration depth of the superconductor.t5 Here again, a determination of the absolute value of /1 is considerably more difficult. In addition, if the superconducting film is thin compared to A, the measurement is sensitive to the effective microwave impedance of the region behind the superconducting film. All of the measurement techniques mentioned above have limitations, particularly in the determination of h ( T) at low temperatures. In measurements of the thin film susceptibility, as well as mutual inductance, resolution suffers when the film is completely shielding; these techniques measure only noise in the system one the film is in the saturated shielding state. The Schawlow-Devlin technique also suffers from nonlinearities in the background at low temperatures, l8 while resonant cavities are limited by their finite sensitivity to the sample inductance. In this article we 1602 Rev. Sci. Instrum., Vol. 62, No. 7, July 1991 describe a simple technique that gives a very high resolution penetration depth measurement, particularly at low temperatures where other techniques fail. The technique is based on measurements of the resonant frequency of a micros@ resonator constructed using superconducting thin films. II. PRINCIPLE OF THE MEASUREMENT As seen above, a simple way to determine the magnetic penetration depth of a superconductor is to measure the inductance of the thin film. The inductance of a superconducting thin film is the sum of two parts, the magnetic inductance and the kinetic inductance. The magnetic inductance reflects the energy stored in magnetic fields created in the region around and inside the film due to currents in the film, and depends on the conductor geometry and the penetration depth. The kinetic inductance reflects the energy stored in the inertia of the superconducting carriers, and is proportional to L2 divided by the cross sectional area of the film, for films thin compared to Lzo12 For suitable thin film geometries,20-23 the kinetic inductance can be made large and very sensitive to Iz. In fact, the present work grew out of the development of a kinetic inductance dominated variable delay transmission line for use at millimeter-wave frequencies.22 A direct measurement of the thin film inductance is difficult. A common technique to measure the inductance is to put the film into an electrical circuit and measure a Dielectric: E, Strip conductor: w thickness d thickness t= b Ground condu thick < I, (a) Patterned strb film I/ v FIG. 1. (a) The microstrip transmission line geometry. The strip conductor has a thickness, tn a linewidth, W, and a penetration depth, A, The ground plane has a thickness, $ and a penetration depth, A, The dielectric has a thickness, d, and a dielectric constant e,fw (b) The fiip-chip implementation of the microstrip geometry, which is realized by clamping a dielectric sheet between an unpatterned ground plane film and a patterned strip film. Superconducting thin films 1802 (b),

3 property of that circuit that depends on the inductance.5 23 to signal and ground plane respectively (Fig. 1). The fac- In this technique we measure the phase velocity of a mi- tors k and g, take the fringing fields and ground/strip crostrip transmission line [Fig. 1 (a)] fabricated using su- asymmetries of the two-dimensional microstrip geometry perconducting thin films. The phase velocity for a signal into account, and are obtained from formulas derived by propagating on a lossless transmission line is simply Chang.24 Note that we have assumed that the ground and VP = (LC) - 2, where L is the series inductance per unit strip superconducting films have the same penetration length and C is the shunt capacitance per unit length. The depth. The capacitance per unit length for a microstrip line inductance per unit length for a microstrip line is given by is given by L=(j.~,#cw)[d+/lcoth(tJ~) +/lcoth(@) +Agccsch(tJA)l, (1) where w is the width of the strip film, t is the film thickness, d is the dielectric thickness, and he subscripts s and g refer I C=eeffeo (kw/d), (2) where the effective dielectric constant, E,~, takes the multiple dielectric system of the microstrip geometry into account.25 Hence the phase velocity for a lossless microstrip transmission line is given by c/ &ii %= Jl + (/l/d)[coth(tj,%) +coth(t&) +g,csch(t,/a)] (3) I where c = ( ec,uo) is the speed of light in vacuum. Note that the phase velocity is independent of the factor k, but still has contributions from field fringing through E,,~ and g, We have also investigated the effect of losses on the phase velocity of a superconducting microstrip line. We find that for typical measurement frequencies (I-10 GHz) and T < 0.98 To the effect of losses on the phase velocity is negligible. A more detailed discussion of this can be found in Sec. V. In practice, time domain measurements of the phase velocity of a transmission line are difficult to perform. However, the phase velocity can be easily measured as the resonant frequency of a finite section of transmission line which is terminated by large impedance mismatches. The resonance condition is that an integral number of half wavelengths span the transmission line s length, L. From this condition we see that the phase velocity is given by v,=2lf,,/n, where f, is the nth resonant frequency. We have chosen Lz4 cm so that the resonant frequencies are in the GHz range. Measuring the phase velocity in the frequency domain brings with it certain advantages. In particular, since microwave frequency sources can be obtained with stabilities of one part in 109, a resonant frequency determination can potentially be done with very great precision. In practice, the resonant frequencies cannot be determined to one part in lo9 because there is finite attenuation in the transmission line, leading to broadening of the resonance peaks. This broadening is proportional to the losses suffered by the microwave signal, so the resolution of the penetration depth measurement is higher for microstrips with lower losses. Consequently, the microstrip resonator method is well suited for measuring n(t) in superconductors at low temperatures where the film losses are lowest. The microstrip resonator measurement has its origins in the work of Pippard26 who first suggested measuring the phase velocity of a signal propagating on a superconduct- ing transmission line to determine the inductance and magnetic penetration depth of the materials. Young et al. were the first to use this technique to measure the penetration depth temperature dependence in thin films.27 The technique was later adopted by Mason and Gould28 to measure the penetration depth and coherence length in indium. Henkels and Kircher29 used microstrip lines to measure the mean free path dependence of the penetration depth in Pb alloys. They also confirmed the thickness dependence of the effective penetration depth for films thin compared to the penetration depth, A,,, = A coth( t/a). A method similar to this, but designed to measure surface resistance rather than inductance, has been successfully used to measure the surface resistance of high temperature superconducting films.30 III. MICROSTRIP RESONATOR MEASUREMENT The resonator measurement setup is shown in Fig. 2. First we will describe the sample probe that allows the microwave measurement of a resonator that is cooled to low temperatures. The probe consists of a 1.5m-long thinwalled stainless steel tube with a probe tip attached to one end and a room-temperature connection box attached to the other. The probe tip is composed of two separable parts as shown in Fig. 2(b). The upper part of the probe tip is a dc terminal breadboard that is rigidly attached to the stainless steel probe tube and houses two single in-line package (SIP) strips, while the lower part of the probe is the sample platform. The pins of the SIPS are directly connected to BNC terminals on the room temperature connection box by means of small gauge copper wires running the length of the probe tube. Detachable electrical connections to the probe tip can then be made from the SIPS. The microstrip transmission line is assembled in a flip-chip geometry [Fig. 1 (b)] directly on the probe tip sample platform as shown in Fig. 2(b). In this geometry, two films are clamped together face to face, with a dielec Rev. Sci. Instrum., Vol. 62, No. 7, July 1991 Superconducting thin films 1803

4 (b) Directional Coupler Connection box (al 0 Microwave Source 2/ (0.01 to 26.5 GHz) Coaxial LOC Termfnal Breadboard Cable Lz;!;$;, Liquid Dewar Resonator He Sample zer Platform rclamp LAlumrnized Teflon FIG. 2. (a) Schematic illustration of the low-temperature microwave measurement setup. (b) Here we see the probe tip in more detail. The flip-chip microstrip resonator is mounted on the sample platform. tric sheet in between. One film is lithographically patterned with a long meandering line terminated with coupling pads, while the other film is left unpatterned to form the microstrip ground plane. We use two films with nominally identical penetration depths for a given microstrip resonator in order to simplify the data analysis. The two films are usually obtained from a larger sample that has been cut into 0.25 in. or 0.5 cm square samples. One could also use a superconductor that has a known penetration depth as the ground plane. We typically use mylar or Teflon FEP3* sheets of 1 or 1 mil thickness, or thin (40-50 pm thick) sapphire wafers for the dielectric. The dielectric sheets have a thin aluminum layer deposited on part of its surface in order to provide a ground contact between the back edge of the ground film surface (the.top film in Fig. 2) and the Cu sample clamp. The dielectric sheet is not metallized over that part of its surface that forms the microstrip line. The flip-chip geometry results in a multiple dielectric medium that complicates the calculation of ee,+ however the measurement results are independent of the dielectrics chosen. We originally fabricated our microstrip transmission lines in this geometry for two reasons. First of all, a monolithic microstrip structure was not possible since high quality multilayer films of high t.emperature superconductors are only now being realized.32 Second, we wished to avoid degradation of the superconduction properties of our films due to excessive processing. Thus, we did not use 1804 Rev. Sci. Instrum., Vol. 62, No. 7, July i991 FEP deposited dielectric layers. One advantage of the flipchip geometry over monolothic structures or deposited dielectrics is that the dielectric thickness can be easily changed. Equation (3) shows that the temperature dependence of the phase velocity increases as the dielectric thickness decreases; thus, smaller dielectric thicknesses are preferred for penetration depth measurements. As explained above, the microstrip transmission line can be assembled directly on the sample platform section of the probe tip (see Fig. 2). However, in a more recent implementation of this measurement technique, the microstrip samples are clamped together in a spring-loaded Cu sample holder, which is then mounted on the probe tip sample platform. This has several advantages over the previous method. First of all, the microstrip assemblies can be removed from the probe tip without breaking the flip-chip stack. Second, the Cu sample holder forms a minimum volume cavity around the microstrip transmission line that serves to raise the resonant frequencies of any spurious package resonances. Third, the springloaded clamping allows better control of strip film to ground film alignment. Last, and perhaps most important, the ground contact is now made via a thin metal sheet that is attached to the Cu sample holder and forms a springlike contact with the edge of the ground plane film. This ground contact is more reliable than the previous method, and ehminates the need to partially metallize the dielectric sheets. The microstrip resonator is excited by microwave signals from a Hewlett Packard 834OB synthesized source, which are brought down to the sample platform by a low thermal conductivity stainless steel coax line (0.086 in. diameter). This same coax line also carries the microwave power reflected from the resonator to a directional coupler, which directs the signal to a Hewlett Packard 8757 Scalar Network Analyzer (SNA). Another identical coax line carries the power transmitted through the resonator from the sample platform to a microwave detector mounted on the top connection box, which directs the signa to the network analyzer. These coax lines are capacitively coupled to the microstrip input pads [Fig. 2(b)]. The lines are free to rotate and move in the longitudinal direction, and their position can be adjusted by moving them through Ultra-Torr fittings at the base of the connection box. This allows adjustment of the resonator coupling. The reflected and transmitted power signals are displayed as a function of frequency on the SNA. In order to obtain the penetration depth of the superconducting films we must obtain reflected and transmitted power traces for temperatures ranging from as low as possible to T,. Temperature control of the resonator can be done in one of two ways. For rapid characterization, the microwave probe can be directly inserted into a helium storage Dewar. The temperature of the sample is then controlled by raising or lowering the probe with respect to the liquid helium surface. However, this method proved to have undesirable side effects at the lowest temperatures due to the helium vapor. As the probe approaches the liquid helium surface, the vapor density changes rapidly, leading to temperature-dependent dielectric effects that can ob- Superconducting thin films 1804

5 scure the effect of the temperature-dependent penetration depth. Because of this, a vacuum can was built to allow measurements in a low background pressure of helium. In this method the probe end is cooled by means of an exchange gas, which is subsequently pumped out. The temperature is then controlled over a broad range by means of a Lake Shore 91C temperature controller with a calibrated Si diode thermometer and a 25 R heater. In either method, electrical connections to Si diode thermometers, heaters, or dc bias leads can be made via the array of BNC connectors on the top connection box. The measurement procedure is as follows. The sample is cooled slowly to the lowest measurement temperature. At this point, the coax lines are adjusted such that the coupling to the resonator is in the loose coupling limit. In this limit both the resonant frequency and resonance Q are independent of coupling strength. Note that loosening the coupling too much results in a reduction of the signal-tonoise ratio, making the measurement more difficult, especially at high temperatures where losses are greater. Experimentally we have observed that a resonator insertion loss (power incident on the resonator minus transmitted power) at low temperatures greater that 20 db is necessary to ensure that the resonator is loosely coupled. The microwave power incident on the resonator is usually set at 0 dbm. When the microwave power is too high, the resonances are degraded by power-dependent losses.33p34 Once the coupling is adjusted, the resonance that we wish to measure is centered in the SNA window, and the temperature is slowly increased using one of the methods outlined above. The reflected and transmitted power traces are transferred to a HP9000 series computer and stored in memory for as many as 500 temperature points. The raw data consist of the reflected and transmitted power traces, PR(f ) and PT(f ), for the temperatures measured. The resonant frequency and resonance Q is obtained from the transmitted power trace, which is fit to a skewed Lorentzian line shape that includes a linear background T3 + T4f PT(f)=T1 + T2f + {I + [(f- T,)2/T,J} by a nonlinear least squares fitting routine. The resonant frequency is f,, = Ts and the resonance Q = T5/2 II?&. For an unskewed Lorentzian ( T2 and T, = 0) we can simply identify f, as the frequency at the peak in the PT(f) curve, and the resonance Q as f,/af3 db, where Af3 db is the difference in frequency between the two half power points of the resonance (relative to the peak power). Once the resonant frequency is determined for all of the measurement temperatures, a plot of phase velocity versus temperature for the microstrip line can be made (Fig. 3). Recall that the phase velocity is very simply related to the nth resonant frequency by the expression, up= 2Lf,,/n, when in the loose coupling limit. Note that since the resonance Q is inversely proportional to line losses, we can also use this technique to determine the microstrip line attenuation. However, obtaining the surface resistance of the superconducting thin films I -b o T/Tc FIG. 3. Typical phase velocity data (normalized to the speed of light) obtained for a YBa$u,O, _ 6 superconducting microstrip resonator plotted vs temperature. by this method is complicated by dielectric and radiation losses, and by the nontrivial analysis necessary to account for the two-dimensional geometry and finite film thickness. Details on the loss analysis of microstrip lines, along with experimental results, will be presented elsewhere.33 IV. DATA ANALYSIS Here we discuss how the temperature dependence and magnitude of the penetration depth are obtained from the phase velocity data. The phase velocity depends on the penetration depth as shown in Eq. (3). In order to obtain ;1 by inverting Eq. (3) one must know the film and dielectric thicknesses as well as the effective dielectric constant of the transmission line, eer. The effective dielectric constant is the most difficult quantity to measure in our implementation of the microstrip resonator technique. Mason and Gould2* and Henkels and Kircher2 were able to determine the dielectric constant directly through measurements of the microstrip capacitance at low frequencies. However, because the thicknesses of the anodized dielectrics used in those studies were on the order of 1000 A, their capacitance values ( nf) are more than three orders of magnitude larger than those of the flip-chip microstrip lines studied here. In the case of the flip-chip microstrip, a capacitive determination of the dielectric constant proved inadequate to allow an accurate determination of the penetration depth. The use of high quality multilayer high-t, superconducting thin films to construct a monolithic microstrip structure should alleviate this problem in the future. Nevertheless, we developed two complementary techniques to analyze the data. One method obtains the temperature dependence of il at low temperatures directly from the measured data, while the other uses a self-consistent fit of the data to an assumed penetration depth temperature dependence to obtain absolute values of /2. I I 1805 Rev. Sci. Instrum., Vol. 82, No. 7, July 1991 Superconducting thin films 1805

6 A. Low-temperature dependence of A(T) We can extract the temperature dependence of ;1( T) at low temperatures from the phlase velocity data without making assumptions about the magnitude of n(t) or imposing a theoretical temperature dependence on the data. The analysis is also independent of the microstrip geometry and dielectric parameter (t,&w,~,~).35-3g Using Eq. (3) we derive the expression ln[ (+d2- (&),i =ln[a(t)coth(&) 2% fin 7, --1(T,)coth(&)] ( ) (4) where T, is some low-temperature baseline temperature. For thick films, such that t > (3/2/2), the hyperbolic cotangent is approximately unity. Th.us, under this assumption and restricting ourselves to temperatures T < TJ2, we obtain ln[ ($g2- (&)2]=lJl(g 1) +b, (5) where b is a nominally temperature-independent constant that depends on /2( T,), Q, t, and d. On the other hand, for thin films, such that t < (A/2), the hyperbolic cotangent can be approximated as (A/t). In this case the first natural logarithm on the right-hand side of Eq. (4) becomes ln{[l*( T) - A2( T,)]/t}, which can be approximated by ln[;l( T)//Z( T,) - I] plus a temperature-independent constant, since the penetration depth does not change much from its low-temperature baseline temperature value for T < TJ2. Thus, JZq. (5) holds for both thin and thick films such that t > (3i1/2) or t < (A/2) in the low-temperature limit, T < TJ2,. For YBazCu307 _ 6, the zero temperature penetration depth is on the order of 1500 A, so this analysis is valid for films with thickness less than 750 A or greater than 2250 A. The YBa2Cu30;l _ 6 films discussed here have thicknesses greater than 4000 A, so Eq. (5) is valid. Equation (5) shows that the phase velocity data can be arranged such that its temperature dependence is solely that of the penetration depth at low temperatures. One implicit assumption is that the geometry, dielectric properties, and losses of the microstrip resonator do not affect the temperature dependence of the phase velocity. This assumption will be discussed in Sec. V. As we will see, the largest problem arises with respect to the temperature dependence of E,~ The low-temperature dependence of/z ( T) simply measures the low-lying pair-breaking excitations that deplete the density of superconducting electrons. The low-temperature dependence of presently known superconductors usually takes one of two forms. lp3 For a superconductor with a finite energy gap over the entire Fermi surface of minimum value A( 0), the low-temperature dependence ( T < TJZ) is given by@ exp[ - A(O)/k,T]. (6) Using Eq. (5) we see that the phase velocity data can be plotted as ln[ (--$&)*- (--&) ]Z(s): +& (7) where b varies slowly with temperature compared to the first term on the right hand side. Since the temperature dependence of Eq. (7) is dominated by the ( f/t) term, a plot of the phase velocity data in the form of Eq. (7) vs TJT will approximate a straight line with slope - A(O)/ kbtc at low temperatures if the superconducting film has a finite energy gap over the entire Fermi surface of minimum value A(0). The baseline temperature in this case is T, = 0 K; v,(o) is obtained by extrapolation from the lowest temperature data (2-4 K). Representative data for low-t, and high-t, films and the BCS theoretical prediction in the weak coupling limit [2A(O)/k,T, = 3.51 is plotted in this fashion in Fig. 4(a). The slopes of the lines in Fig. 4, as well as their shape, are only slightly sensitive to the choice of v,(o). Another observed low-temperature dependence for A( T) in superconductors is that associated with nodes in the superconducting energy gap, and has the form 13 Using Eq. (5) we see that the phase velocity data can be plotted as ln[ ($) - (--&)i]=nln(~) +const. (9) In this case, a plot of the phase velocity in the form of Eq. (9) vs In( T/T,) will approximate a straight line with slope equal to the power law exponent, n, at low temperatures. To test for this possibility, the data in Fig. 4(a) are replotted in Fig. 4(b) over the same temperature range, but now as a function of ln( T/T,). 6. Absolute penetration depth While the analysis above is very useful in determining the low temperature dependence of il( T) without fitting the data, it does not tell us anything about A at higher temperatures, nor does it give us absolute values of /2. The absolute value of il over the entire temperature range can be obtained by inverting Eq. (3) to obtain the penetration depth from the phase velocity data once the effective dielectric constant of the microstrip line is determined. However, as explained previously, we are unable to directly measure the effective dielectric constant with sufficient accuracy. Analytic expressions to calculate eeff are also not very useful since these formulas are typically only accurate to about 5%.25 Because of these difficulties, we have developed an analysis technique that allows us to obtain Q from the phase velocity data itself by self-consistently fitting the data to an assumed penetration depth temperature (8) 1808 Rev. SC!. instrum., Vol. 82, No. 7, July 1991 Superconducting thin iilms 1806

7 where we have written the penetration depth in terms of its zero temperature value /lo and a normalized penetration depth Z( T/T,) =a( T/T,)/&. We have also defined the factor (b) 0 \ TclT In (TiTc) FIG. 4. (a) The phase velocity data for Nb, NbCN, and YBa2Cu307 _ 6 films is plotted vs TJT, as suggested by Eq. (7), for reduced temperatures from 0.25 to 0.5. The low-temperature slope of these plots can be interpreted as the energy gap of an exponentially activated A(T). The result obtained for weak-coupled BCS theory is also shown (Ref. 42). (b) The same data now plotted vs ln( T/T,), as suggested by Eq. (9), for the same temperature range. The low-temperature slope of these plots can be interpreted as the exponent, n, of a power law A(T) temperature dependence. Note that both plots are normalized at T/T, = 0.5. dependence. Note that the deduced absolute values of I. are strongly dependent on the form of the temperature dependence assumed. This difficulty is not unique to our measurement technique; for example, dc magnetization measurements rely on similar fits to obtain absolute values of k4 We will briefly outline the procedure here; a more detailed description can be found in the following references.21v33 Equation (3) can be rearranged as (&J=4r+ (~)a,2 (;)Mm (10) M(T)=f[coth( 6) +g,csch( &j) G + Oth 01 /z(t) (11) which takes the two-dimensional microstrip geometry and finite film thicknesses into account and allows us to linearize the equation and simplify the computation. The phase velocity data are plotted as (c/q2 vs Z( T)M( T) for a given T,, where the normalized penetration depth Z(T) is theoretically calculated. Since M(T) depends on the unknown R values as shown in Eq. ( 11)) we initially assume that M(T) = 1 for all T. Equation (10) shows that the slope of a linear least square fit of the data when plotted in this manner is proportional to &, while the intercept is equal to E,~ Once eeff is determined, Eq. (3) can be inverted to directly obtain penetration depth values. The factor M(T) can now be calculated using these penetration depth values and measured film and dielectric thicknesses. The above steps then are repeated until M(T) converges. At this point the entire procedure is repeated interactively, varying T, until the best linear least square fit is achieved. T, values obtained in this manner agree very well with those obtained from dc resistivity measurements. The analysis procedure outlined above results in il( T) values calculated from the phase velocity data using Eq. (3) and the effective dielectric constant value obtained from the best fit. We call this the data penetration depth curve. The procedure also results in I( T) values obtained by multiplying the zero temperature penetration depth value obtained from the best fit, &, by the assumed normalized penetration depth Z(T). We call this the fitted penetration depth curve. When the assumed temperature dependence fits the data very well, these two penetration depth curves will be identical. Conversely, if the assumed temperature dependence does not fit the data very well, these two curves can be quite different. The specific procedure that we have developed to obtain the absolute value of,i( T) over the entire temperature range is as follows. First, the data are fit to an assumed temperature dependence Z(T) calculated using weak-coupled BCS theory (taking the energy gap 2A(O)/k,T, as a variable, while the electronic mean free path I, the coherence length go, and the London penetration depth ill are determined independently).42 We then change the form of Z(T) by adjusting the energy gap parameter until the data and fit penetration depth curves approach each other. Egloff, Raychaudhuri, and Rinderer43 used BCS theory with the energy gap as an adjustable parameter to obtain accurate penetration depth values for lead. However, they did not use self-consistent fits of the theory to the penetration depth data in order to obtain the energy gap, but used gap values obtained from tunneling measurements. The procedure of varying the energy gap within 1807 Rev. Sci. Instrum., Vol. 62, No. 7, July 1991 Superconducting thin films 1807

8 weak-coupled BCS calculations to model strong coupling superconductors has also been validated by Blocksdorf and V. MEASUREMENT RESOLUTION AND SYSTEMATIC ERRORS The main advantage of the microstrip resonator technique over other methods of determining the penetration depth is that it is. an extremely high resolution measurement. In this section we will quantify the resolution of our penetration depth measurements. We will also look at systematic errors that can arise in the penetration depth measurement. These errors are mainly due to the presence of losses, and thermal variations of the microstrip geometry and dielectric properties. In order to look at the resolution and systematic errors in the determination of the penetration depth from the phase velocity data, we need to invert Eq. (3) and obtain an expression for il in terms of measured quantities. Since most of our measurements are clone using resonators that are fabricated in the wide line (w>d) and thick film (t > 3/2/2) limits, we will investigate the measurement resolution and systematic errors in this limit. For wide lines k = 1 and g, = 0, while for thick films coth( t/a) = 1. With these assumptions, we can invert Eq. (3) to obtain: a=;[ &( $-11. (12) It can be shown that for narrow lines and/or thin films the measurement resolution will increase and the magnitude of systematic errors will decrease since the line inductance for a given value of ;1 is increased in those limits.33 Thus, using Eq. ( 12) for the measurement pathology analysis will give a conservative estimate of both measurement resolution and systematic errors. A. Measurement resolution The resolution in,l is limlited by the resolution to which changes in the phase velocity can be measured. An estimate of the minimum resolvable change in the phase velocity, &+,, can be made as follows. The phase velocity is determined from the resonant frequency of the transmitted power versus frequency data. The transmitted power data consists of 801 discrete frequency data points with the resonance peak centered in a frequency window with a width chosen to be approximately 3Aj ;,a. Recall that Af3dB is related to the resonance quality factor by Q = f,/af 3 da. Thus, the minimum resolvable fractional frequency change is W 3AfsciB 1 y=801f=iz@ The minimum resolvable fractional phase velocity change is equal to the minimum resolvable fractional frequency change since up is linearly related to f,,,. A conservative Q value for a typical microstrip resonator (dielectric thickness d = 0.5 mil) is 1000, which gives a fractional up change of approximately Sv/vP = 4 X Using Eq Rev. Sci. Instrum., Vol. 62, No. 7, July 1991 ( 12) it can be shown that the minimum resolvable penetration depth change corresponding to Sv, can be written as&? z d( Su&,). Using the numbers above the translates to a conservative penetration depth measurement resolution of approximately 0.5 A. This result is essentially independent of the dielectric thickness since Q K d. Thus we see that the microstrip resonator measurement allows a very high resolution measurement of the penetration depth, which is especially important at low temperatures where an exponentially activated penetration depth does not change very much with temperature. B. Transmission line losses When the resonator is Iossy, the resonator frequency of the microstrip line can be pulled down by the finite resonance Q. To see this, consider the familiar equation of motion for a damped harmonic oscillator which is sinusoidally driven, or mjt + /3k + kx=f X + 2A.t + &=A cos(wt) cos(ot), where wg = k/m is the undamped oscillation frequency and the damping parameter is 2A = f3/m. If the damping is light enough (A < wc), the system will undergo damped oscillations with an oscillation frequency reduced from its undamped value, de, = co; - 2/x2. In terms of the oscillator Q = w,/2/2, quency in the presence of damping is the oscillation fre- c~$~=ci$[ 1 + (l/2@)] - or w,~g~~[ 1 - (l/4@) 1. Thus the fractional change in the oscillation frequency due to losses is Such a shift in resonant frequency is significantly less than the measurement resolution with practical Q values (Sf/fo ;= l/2670). Hence we can neglect the effect of finite losses on the resonant frequency. The only exception to this conclusion will be near TO where losses may lead to Q values less than 50. However, this will influence the determination of the penetration depth only for temperatures T > 0.98TC.33 The above analysis assumes that the inertial term m of the harmonic oscillator (analogous to the kinetic inductance of the microstrip resonator) is not affected by the presence of damping. This is true in superconducting transmission lines, at least up to signal power levels for which the film inductance does not change.34 At higher power levels the film inductance will increase because of current induced depairing or flux entry, but these phenomena occur beyond the power levels used in the measurements discussed here. However, in normal metal transmission lines, the situation is very different because the skin depth Superconducting thin films 1808

9 (and hence the inductance) depends directly on the conductivity.45 One finds that in normal metal transmission line 1 2--, 00 2Q and the resonant frequency is significantly lowered. C. Thermal variations of transmission line parameters As seen in Eq. (12), the penetration depth value depends on the dielectric thickness d and the effective dielectric constant E,~ In this section we will look at the effect of temperature-dependent dielectric properties. Note that in the thick film (t> 3U2) limit the penetration depth is independent of the film thickness variations. If we assume deviations in the dielectric thickness and effective dielectric constant of 6d and 6~ respectively we can derive a fractional deviation in,l using Eq. ( 12):!$(y)-;(E$). (13) However, before moving on to temperature dependencies, simply look at how a temperature-independent error in the determination of d or E,.~ affects the penetration depth values obtained. The dielectric thickness d is set by the thickness of the dielectric sheets, which is measured with an accuracy of approximately f 5%. This implies a fractional deviation of f 5% in the absolute value of the penetration depth (for YBa#&O,-6 d-1500 A, and &l.-75 A). However, since 6il scales with il, a temperature-independent error in d will not affect the determination of the temperature dependence, A( T)/;1,, of the penetration depth. An error in the determination of the effective dielectric constant is more problematic for two reasons: ( 1) the prefactor (d/w) in the above equation can be very large at low temperatures ( -SO), and (2) the error in the penetration depth, S/z, does not scale with il, which implies that the temperature dependence as well as the absolute magnitude of A will be affected. This is why analytical formulas using tabulated dielectric constant values (accuracy - 5%) or measurements of the low frequency capacitance cannot be used to determine the effective dielectric constant with sufficient accuracy to allow the determination of the penetration depth. Next consider the effects of a temperature-dependent dielectric thickness d, film thickness t, or effective dielectric constant E,~ These variations will have an affect on il given by Eq. ( 13). Let us consider a temperature dependent dielectric thickness first. The low-temperature thermal expansion coefficient of Teflon FEP is a x 10-5/ K3 which has the largest thermal expansion coefficient of any of the dielectrics used in the resonators.33 The cumulative error in the penetration depth over a measurement temperature range AT is given by 61 z &,aat. For YBazCu 0, _ a (A, , and AT-90 K) this implies a/2-2 8, while for Nb (ilo-750 8, and AT-9 K) 6A -0.1 A. Since the measurement resolution is on the order of 1 A or less, the effect of a temperature-dependent dielectric thickness can be ignored. Although the penetration depth analysis is independent of film thickness in the thick film limit, changes in the film thickness have a similar effect on the penetration depth as the dielectric thickness if we use the relevant equations for the thin film limit. The thermal expansion coefficient of YBa2Cu30,-6, and other typical superconducting materials, at low temperatures should be much less than its room temperature value of approximately 10-5/K.46 This expansion coefficient leads to cumulative &l values similar to those shown above, and can therefore be ignored. Finally, the effect of a temperature-dependent effective dielectric constant is considered. Because of the two-dimensional field distribution in the microstrip geometry (e.g., see Fig. 7 of Ref. 47), electric fields are not confined entirely to the dielectric between the strip and ground plane films. Thus the effective dielectric constant depends on the temperature-dependent dielectric constants of both the sheet dielectric and the substrate on which the strip conductor is deposited. For typical resonator geometries (w/d-5-10) and substrates such as MgO (E, - lo), LaA103 (E,- 25 ), yttria-stabilized zirconia (YSZ) (e,-25), or sapphire ( et- lo), which have dielectric constants significantly larger than that of the sheet dielectric (for Teflon, E, - 2), analytic formulas2 indicate that the temperature dependence of the substrate should dominate the temperature dependence of E,~ This assumes that the temperature dependence of E, for Teflon (we have no data for temperatures below 100 K) is not much larger than that of the substrates. We have estimated the value of he/e, at low temperature for the above substrates: MgO, LaAIO,, and sapphire (he/e, < 1 X 10-5/K),48*49 and YSZ ( AE,/E, - 2 X 10-4/K).48,49 The estimate for MgO, LaA103, and sapphire is believed to be an upper limit since the e,(t) curves for these materials become very flat at low temperatures. The magnitude of Ae,./e, for YSZ is much larger than that of the other substrates, and stays relatively constant up to K.49 Figure 5 shows the resonant frequency data for a microstrip resonator fabricated using YBa2Cu30, _ 6 films on YSZ. The temperature dependence of the resonant frequency is clearly dominated by the linear temperature dependence of YSZ at low temperatures. The slope of a linear fit to the data at low temperatures ( 1.05 X 10-4/K as shown in Fig. 5) can be used to estimate AE,/E, for YSZ. If the resonant frequency change is due solely to e,(t) of the film substrate, then Af,Jf res = Au/up = 0.5 he/e,, and the fit obtained from the f,,(t) data implies AeJer = 2.1 X 10-4/K for YSZ. This is in very good agreement with independent measurements.49 Obviously, superconducting films deposited on substrates with very highly temperaturedependent dielectric constants, such as YSZ or SrTi03,15 are not well suited for 2 ( T) measurements using this technique. The effect of dielectric temperature dependencies on measurements using films deposited on MgO, LaA103, or sapphire is much less, as seen in the data for a resonator fabricated using YBa2Cu,07 _ s films on MgO (Fig. 5) Rev. Sci. Instrom., Vol. 62, No. 7, July 1991 Superconducting thin films 1609

10 s F g T WI FIG. 5. The low-temperature resonant frequency data (normalized to its extrapolated zero temperature value) is shown here for two superconducting resonators. The data for the resonator fabricated using YBa,Cu30, _ 6 films on YSZ clearly shows the effect of the temperaturedependent dielectric constant of YSZ, while the data for the resonator fabricated using YBa,Cu30, _ s films on MgO show no temperature-dependent dielectric effects. We can estimate the magnitude of the cumulative change in penetration depth over a measurement range AT using 6/2 < (1 X 10-5/K) d AT/2, which is obtained from Eq. (13). For YBa2Cu3107-s (d pm and AT- 90 K) this implies S/z < 30 A, while for Nb (d pm and AT-9 K) ail< 3 A. The effect of a temperature dependent e,tf on low T, films is not very large due to the small measurement temperature range, as is borne out by the fact that measurements of the temperature dependence of the penetration depth of Nb and NbCN films on sapphire substrates are in very galod agreement with theory (see the following section). On {the other hand, the effect of a temperature-dependent eeff on high T, films such as YBa2Cu307 _ 6 is larger, although the estimate above is an upper limit. To put a cumulative penetration depth deviation over the full temperature range of -30 A into perspective, note that a conventional weak-coupled BCS superconductor with /2, A, 2A(0)/k~T, = 3.5 and T, = 90 K has a theoretically calculated42 penetration depth which changes by more than 2800 A from T = 0 to 85 K. We have also conducted detailed systematic experiments that verify that the change of the resonant frequency with temperature of our YBa2Cu307 _ 6 resonators is consistent with n(t) being the only temperature-dependent parameter.33 This gives us confidence that our penetration depth measurements are not being afffected by temperature-dependent dielectric properties, or any other parasitic temperature dependence Rev. Sci. Instrum., Vol. 62, No. 7, July 1991 Vi. EXPERIMENTAL RESULTS In this section we will present some experimental results obtained using the microstrip resonator technique. Since the focus of this article is the measurement technique itself we will be brief; a more detailed account of the results of our penetration depth measurements of both high- and low-t, films can be found in Refs. 21, 33, and The results presented here were obtained for Nb, NbCN, and YBa2Cu307 _ s thin films. The Nb films used were sputtered5 on sapphire l-102 substrates, and had low residual resistivity ratios ( z4), but high T, ( K). These films were 3650 A thick. The NbCN films were prepared at TRW by reactive sputtering on sapphire substrates,50 were 5660 A thick, and had T, - 16 K. YBa2Cu307 _ 6 films were prepared at Stanford by in-situ off-axis sputtering5r on MgO, YSZ, and LaAlO, substrates. These YBa2Cu307 _ 8 films are c-axis normal oriented and heavily twinned, implying that the shielding currents used to measure the penetration depth flow roughly equally in the a and b directions. These films also show high critical current densities52 and good rf properties.53*54 The YBa2Cus07-h films used in the measurements presented here were sputtered to a thickness of 4340 L% on MgO, and had T,- 82 K as determined from dc resistivity measurements. Microstrip resonators were fabricated using these superconducting thin films. The width, w, of the patterned strip conductor was 103,75, and 47 pm for the Nb, NbCN, and YBa2Cu30, _ 6 thin films, respectively. All of the resonators used a 12.7~,um-thick Teflon FEP3 sheet as the dielectric, and were measured using the experimental procedure outlined in Sec. III. Representative phase velocity versus temperature data is shown in Fig. 3. The low-temperature dependence of /2(T) can be investigated directly, as discussed in Sec. IV A. When the phase velocity data is plotted as suggested by Eq. (7)) the exponential temperature dependence should make the data fall on a line with slope - A(O)/k,T,. Figure 4(a) shows that the films follow this low-temperature exponential de- pendence to a very good approximation. The plots for the Nb and NbCN films show an inferred energy gap of 2A(0)/kBT, = 4.4 f 0.5 and 4.8AO.5, respectively, while the plot for the YBa2Cu307 _ 6 film shows an inferred energy gap of 2A(O)/kBT, = 2.5 f 0.2. We are more confi- dent about the value deduced for YBa2CuJ07-6 since those data extend down to lower reduced temperatures ( T/T,) than do the NbCN or Nb data. In Fig. 4(b) the phase velocity data are plotted as suggested by Eq. (9), which allows one to check for power law behavior in the low-temperature penetration depth. The data for all of the films (as well as BCS weak-coupled theory) show a slight but distinct downward curvature, which is consistent with what we would expect for an exponentially activated penetration depth. If we ignore this curvature, we can obtain values of the power law exponent for the temperature dependence of 2 on the films: for the Nb film II - 6.0, for the NbCN film n-6.5, and for the YBCO film, n Of course, given that the data are better fit by an exponential temperature dependence, the physical significance of these Superconducting thin films 1810